JAMA Netw Open. 2022 Aug 1;5(8):e2225964. doi: 10.1001/jamanetworkopen.2022.25964.
IMPORTANCE: Instrumental variables can control for selection bias in observational research. However, valid instruments are challenging to identify.
OBJECTIVE: To evaluate regional variation in sleeve gastrectomy following insurance coverage implementation as an instrumental variable in comparative effectiveness research.
DESIGN, SETTING, AND PARTICIPANTS: This serial cross-sectional study included adult patients in a national Medicare claims database who underwent sleeve gastrectomy or Roux-en-Y gastric bypass from 2012 to 2017. Data analysis was performed from January to June 2021.
EXPOSURES: Laparoscopic sleeve gastrectomy and Roux-en-Y gastric bypass.
MAIN OUTCOMES AND MEASURES: The association of the instrumental variable with treatment (ie, undergoing sleeve gastrectomy), as well as mortality, complications, emergency department visits, hospitalization, reinterventions, and surgical revision.
RESULTS: A total of 76 077 patients underwent bariatric surgery, of whom 44 367 underwent sleeve gastrectomy (mean [SD] age, 56.9 [11.9] years; 32 559 [73.5%] women) and 31 710 underwent gastric bypass (mean (SD) age, 55.9 (11.8) years; 23 750 [74.9%] women). After insurance coverage initiation, there was substantial regional and temporal variation in adoption of sleeve gastrectomy. Prior-year state-level utilization of sleeve gastrectomy was highly associated with undergoing sleeve gastrectomy (Kleibergen-Paap Wald F statistic, 910.3). All but 2 patient characteristics (race and diagnosis of depression) were well-balanced between the top and bottom quartiles of the instrumental variable. Regarding 1-year outcomes, compared with patients undergoing gastric bypass, patients undergoing sleeve gastrectomy had a lower 1-year risk of mortality (0.9%; 95% CI, 0.8%-1.1% vs 1.7%; 95% CI, 1.3%-2.0%), complications (11.6%; 95% CI, 10.9%-12.3% vs 14.1%; 95% CI, 13.0%-15.3%), emergency department visits (48.3%; 95% CI, 46.9%-49.8% vs 53.6%; 95% CI, 52.3%-55.0%), hospitalization (23.4%; 95% CI, 22.4%-24.4% vs 26.5%; 95% CI, 25.1%-28.0%), and reinterventions (8.7%; 95% CI, 8.0%-9.4% vs 12.2%; 95% CI, 11.2%-13.3%). The risk of revision was not different between groups (0.6%; 95% CI, 0.3%-0.8% vs 0.4%; 95% CI, 0.3%-0.6%).
CONCLUSIONS AND RELEVANCE: In this cross-sectional study of patients undergoing bariatric surgery, there was significant geographic variation in the use of sleeve gastrectomy following initiation of insurance coverage, which served as a strong instrument to compare 2 bariatric surgical procedures. This approach could be applied to other areas of health services research to serve as a complement to clinical trials.